7

I am familiar with the xts subsetting abilities. However, I can't find an elegant way to subset a parameterized range of dates. something like this:

times = c(as.POSIXct("2012-11-03 09:45:00 IST"),
          as.POSIXct("2012-11-05 09:45:00 IST"))

#create an xts object:
xts.obj = xts(c(1,2),order.by = times)

#filter with these dates:
start.date = as.POSIXct("2012-11-03")
end.date = as.POSIXct("2012-11-04")

#instead of xts["2012-11-03"/"2012-11-04"], do something like this:
xts[start.date:end.date]

Does anybody have any idea? Thanks!

zuuz
  • 859
  • 1
  • 12
  • 23

4 Answers4

9

You could paste the start.date and end.date objects together, separating by "::" or "/", and then use that to subset.

R> xts.obj[paste(start.date,end.date,sep="::")]
                    [,1]
2012-11-03 09:45:00    1
Joshua Ulrich
  • 173,410
  • 32
  • 338
  • 418
  • I hesitate to put this but when I read the help I thought it is more efficient to give timeBased index. isn'it? – agstudy Dec 31 '12 at 13:28
  • @Joshua: thanks a lot! [even though the way you put, it suddenly looks trivial :-)...] – zuuz Dec 31 '12 at 13:36
  • @agstudy: ISO-style character subsetting _by range_ (i.e. a length 1 vector containing `:` or `/`) is very fast. It is slow is if you subset with an ISO-style character _vector_. – Joshua Ulrich Dec 31 '12 at 13:37
  • @JoshuaUlrich: actually, if your end.date = as.POSIXct("2012-11-05") [instead of the 2012-11-4], then using sep=":" returns only the 1st element. You actually need sep="/". Should I edit your answer? – zuuz Dec 31 '12 at 13:51
  • @zorbar: whoops, that should have been `::`. – Joshua Ulrich Dec 31 '12 at 13:52
3

from the help of [.xts {xts}

As xts uses POSIXct time representations of all user-level index classes internally, the fastest timeBased subsetting will always be from POSIXct objects, regardless of the indexClass of the original object.

So you can do subsetting timeBased like this :

xts.obj[seq(start.date,end.date,by=60)]
                    [,1]
2012-11-03 09:45:00    1
agstudy
  • 119,832
  • 17
  • 199
  • 261
  • Thanks for you answer. however, start.date and end.date are POSIXct objects. thus, a sequence like you write gave me a sequence with 1-second increments. So, I used it as xts.obj[seq(as.Date(start.date),as.Date(end.date),by=1)] but it gives me an empty list... – zuuz Dec 31 '12 at 13:26
1

I just needed to do the same thing. Here is my solution, based on the original example.

library(xts)

times = c(as.POSIXct("2012-11-03 09:45:00 IST"),
          as.POSIXct("2012-11-05 09:45:00 IST"))

#create an xts object:
xts.obj = xts(c(1,2),order.by = times)

#filter with these dates:
start.date = as.POSIXct("2012-11-03")
end.date = as.POSIXct("2012-11-04")

# By using an index that is the logical AND of two vectors
xts.obj[start.date <= index(xts.obj) & index(xts.obj) <= end.date]
1

For those who are racking their brain to do this for non Posix. Especially for quarter based data i.e. 2001 Q2 to 2006 Q3.

I used a simple yet elegant solution:

library(xts)

starting.quarter<-"200101"
ending.quarter<-"201702"

oil_price_by_qtr<-oil_price_by_qtr[paste(starting.quarter,ending.quarter,sep="/")]

And this will subset the XTS object from 2001 Q1 to 2017 Q2.

May this help some other poor soul avoid losing 2 hours of his life.

Unheilig
  • 16,196
  • 193
  • 68
  • 98